Easing the Cognitive Load: Where AI Supports the Human Work of Care

AI could make a world of difference in healthcare in a myriad of areas, but there is one place where it is desperately needed and seemingly unnoticed. Clinician burnout caused by the daily struggles to manage inefficiencies in the healthcare system is silently driving good clinicians away. The first major use of AI should be here. Clinicians need support. If they are burnt out and depleted, no amount of tech will help them.

Burnout Is a Design Problem

Burnout is often thought of as an individual problem, like a clinician needs to improve their boundaries or they need to go to a wellness retreat. That may even be an easy answer. But, no. It is an ineffective system. Clinicians are now forced to spend time clicking through a poorly structured digital system, in addition to being a clinician, to be a coder, a message responder, and an approval chaser. It is no wonder many clinicians have lost the passion for being a clinician. Systems need to be designed with the clinician in mind or else they lose the primary purpose of being a clinician.

The system we have asked them to work inside is a mess. That is why administrative software has taken precedence over the care showing a complete lack of respect for the needs of the clinicians. This system has to change. It is entirely a broken system we have asked them to operate in.

What Cognitive Load Actually Is

Cognitive load is difficult to quantify as it is a concept lacking in defined symptoms. It accounts for the pressure of having so many unfinished tasks that your attention is constantly pulled. This is especially true when your attention must be divided between your patient and your keyboard and the various tasks associated with patient care when your email and/or task list refuses to pause. No single element contributes to the cognitive load, but taken together, it contributes to general fatigue and cognitive depletion.

This monotony of switching and refocusing on different tasks takes an expensive and invisible toll. This is true of most disruptions; a small effort is needed each time to get back to the task you left. Hundreds of these disruptions build an omnipresent cognitive load. This is the effect most modern "productivity" tools have. Every new system is supposed to help, but really replaces previous systems with its own set of demands, logins, and notifications. This burden on the clinicians has only increased as these systems have been developed and deployed while the official story kept insisting on better productivity.

Identifying Burnout

The word "burnout" is often used without precision, but defining its breadth is important. Part of it is simply exhaustion from long hours with insufficient rest, but there is more to it than the tiredness. It is hard to describe the rest. Clinicians know what good patient care is. They spend most of their workdays being prevented from providing patient care by the systems designed to help them. The gap between the patient care they are trained to provide and the patient care the system allows is an injury unto itself. It is the distress they feel when the system and its workflow unjustly prioritize the clerical work of an in-box over patient care. The system and its workflow compel them to divide their attention between a patient, whom they can only afford to give half of their attention, and a computer. Hours spent in this manner without seeing the results of their efforts for the good of the patient is what drives the clinician to walk away.

Areas Where Technology can Assist

When used properly, AI can provide genuine assistance. Ambient transcription can be used to make notes during a patient care session without the transcription being done hours after the session is ended. AI to summarize can help in pulling out important information from records without the clinician having to reconstruct care histories by writing them. Decision support systems that are properly designed for the workflows of clinicians can draw and present the right information when and where it is needed without having to present one more alert. A good in-box system can aid the clinician in prioritizing the many demands on their time and can present canned responses to routine messages without pulling attention away from the patient he is currently seeing.

By themselves, these tools do not eliminate clinician burnout. Individually, they help mitigate burnout only when deployed alongside clinician voice and monitored with responsibility. When tools are designed for users without input from the users, more steps are added or the process is more cumbersome instead of streamlined.

The Trap of Squeezing More

Clinician burnout is usually an underreported adverse effect of developing and incorporating more and more technology into healthcare. An overemphasis on technology to drive higher productivity or to do more in order to force one more patient interaction into the schedule is a clear sign that the purpose of technology in the workplace is being violated. AI and other answers to clinician burnout then become just a more advanced version of the same extractive productivity tool. The time they free will be demanded back before the clinician feels any relief.

These tools have to pass a simple test, whether the day became more humane rather than only more demanding. If the time that AI gifted is reclaimed the moment it is demanded, the tool does not solve the problem the organization set out to solve, employee burnout. The goal has to be to reclaim that time to care for clinicians and other stakeholders, lest it be harvested the moment it reappears.

Build Around the Clinician

The goal of the following principles challenges the prevailing order of system design. Traditionally, clinicians have been expected to adapt to the technology, learning and accommodating the quirks and friction of the system. Adapting technology to the clinician is the preferred option, however, system design will often require the tool to exist within an already established framework of work, rather than creating a new one. To design and implement a tool, a decision must be made with clinical, operational, and technical leadership teams. Each team must participate and equally share responsibility.

To illustrate the principle at hand, consider a system feature that triggers an alert for every value that falls on the border of some clinical threshold. Clinicians will quickly learn to ignore the alert, which is the purpose and functionality of the system. However, if the system were designed to understand context, and therefore discriminate among values that warrant a rare alert, the system would become more valuable to the user. With the same technology, but considering the clinician user during the design phase, the same system ends up either harmful or supportive.

What We Are Actually Protecting

Burnout originates from systems unable to support the humans performing the work. If we want AI to improve care, we need to ensure it first makes the work easier for the humans providing care. AI does not need to be faster or more advanced. It needs to improve the work in ways the humans providing the care can recognize and appreciate.

Care does not happen in one continuous block of time. Care is made up of small moments in between care visits and the moments spent looking at a care recipient before deciding the next steps. These moments should never be seen as waste and should be left untouched. A tool that optimally fills the time that the humans providing care are not working is a tool that misses the point. The goal is not to push care providers to the limits of what they can do with the technology. The goal is to create a system that actually supports the care providers.

Christopher Hutchins Founder and CEO, Hutchins Data Strategy Consultants